As someone who has never written an application to be ran in a cloud environment, I've been wondering for a while now: what sort of extra complexity is introdu

As someone who has never written an application to be ran in a cloud environment, I've been wondering for a while now: what sort of extra complexity is introduced in your code to handle the unreliability of any given server, the network, and latency in general? What techniques are used to deal with it? I'm approaching this more from the viewpoint of a webapp that might only need ~10 dedicated servers to handle peak load with redundancy rather than ~150, so our colo solution doesn't have servers constantly breaking on us.